AreAlmostEqualUlps throws OverflowException for -0

[markhobson]

Similar to #1046, the following assertion:

Assert.That(-0d, Is.EqualTo(0d).Within(1).Ulps);

Results in:

System.OverflowException : Negating the minimum value of a twos complement number is invalid.
   at NUnit.Framework.Constraints.FloatingPointNumerics.AreAlmostEqualUlps(Double left, Double right, Int64 maxUlps)
   at NUnit.Framework.Constraints.Numerics.AreEqual(Double expected, Double actual, Tolerance& tolerance)
   at NUnit.Framework.Constraints.Numerics.AreEqual(Object expected, Object actual, Tolerance& tolerance)
   at NUnit.Framework.Constraints.Comparers.NumericsComparer.Equal(Object x, Object y, Tolerance& tolerance, Boolean topLevelComparison)
   at NUnit.Framework.Constraints.NUnitEqualityComparer.AreEqual(Object x, Object y, Tolerance& tolerance, Boolean topLevelComparison)
   at NUnit.Framework.Constraints.EqualConstraint.ApplyTo[TActual](TActual actual)
   at NUnit.Framework.Assert.That[TActual](TActual actual, IResolveConstraint expression, String message, Object[] args)
   at NUnit.Framework.Assert.That[TActual](TActual actual, IResolveConstraint expression)

Tested against NUnit 3.12.0.

[Poimen]

Ah, the good ol' 0 == -0 problem.

For those interested, there is a good write up on floating numbers here:
https://ciechanow.ski/exposing-floating-point/

The long and short of it is that IEEE floating point numbers represent 0 as two different amounts so 0 and -0 are valid and different, while at the same time 0 == -0 is valid. Math is hard 🤣

@mikkelbu the easier fix is to do a left == right comparison when the signs mismatch to prevent the Math.Abs from throwing. If you would like, I can put a PR together if nobody is looking at this?

[mikkelbu]

@Poimen You are most welcome to look at this problem. Please note that the following is working

Assert.That(-0d, Is.EqualTo(0d));

It is only when we compare the values with tolerance using "Ulps" - unit in the last place - that the problem occur.